Field of the Invention
The present disclosure relates to a technique for generating a color table of color information about a predetermined object.
Description of the Related Art
Recently, the study of mixed reality (MR) has been conducted. In the MR, information relating to a virtual space is superimposed on a physical space in a real time manner to provide it to a user. In the MR, a composite image is displayed, in which computer graphics (CG) is superimposed on a whole or a part of an actual image captured by an imaging apparatus such as a video camera. This CG is an image in the virtual space created according to the position-and-orientation of the imaging apparatus.
At that time, by extracting a hand area from an image in the physical space, an interference determination between a hand in the physical space and the CG or front-and-back relationship determination between the hand and the CG. However, since the shape of the hand is complicated and differs among individuals, it is difficult to detect the hand area correctly and in high speed from an image.
At present, examples of methods for detecting a hand area include a area segmentation method such as a graph cut, a dynamic contour extraction method such as snakes and LevelSet method, and an learning-base extraction method such as a model fitting and feature extraction. In any of the methods, an initial hand area may be determined by extracting a skin color in many cases. Therefore, it is important to register an appropriate skin color database (color table) to achieve high accuracy.
There are various proposals about a hand extraction method using skin color. For example, Japanese Patent Application Laid-Open No. 2005-228140 discusses a method for generating a skin color model by capturing an image of a hand area, then specifying an object area and non-object area on this captured image via a user interface (UI) such as a mouse operation. This method enables generating a skin color model suitable for the current environment (object and environment) easily, without generating a skin color model by using a plurality of images captured by repeating capturing images of many persons.
However, the method discussed in Japanese Patent Application Laid-Open No. 2005-228140 needs to repeat capturing the images and the UI operations a plurality of times in order to obtain a skin color model with sufficient accuracy. As a result, users have to perform extremely cumbersome user operations. In particular, in a case where a color of an object changes depending on where the object is in the captured image due to an influence of peripheral darkening caused by a lens and an image sensor, it is necessary to repeat user operations many times while changing the position of the object in the captured image if a model including this phenomenon is to be created.